How to best provide online usage statistics to content subscribers? Should I?

September 5, 2008 4:38 AM

Tools for a scientific publisher to provide usage statistics for its subscribers? Is this a good idea?

I am the webmaster for a small scientific publisher. Our content is presented online in a way that is similar to most other publishers. The abstracts are free to all, but the articles are restricted to subscribers by the use of IP addresses in htaccess files. Most of our subscribers are university libraries and research institutes.

Lately we have been getting many requests for usage statistics from our subscribers. I have a way of doing this using summary.net, a web log analysis program, but it is time consuming and not an elegant or simple solution requiring me to change the software's settings for each subscriber, render the log files, output an excel file, reformat excel file and then change the software's settings back for my needs. If we decided to offer this service, using this method would take up too much of my time.

Is there some sort of software that I can host on our server that the user can use to show them download statistics for certain files in certain directories based on a list of IP addresses or ranges that they enter?

I thought that a program like awstats might be able to get me in this direction but I don't have root access to our managed server in order to install it properly.

I have also looked at the Project COUNTER as a possible solution for providing statistics, but it seems to be needlessly complicated for what we are trying to accomplish.

Also, is there a way for these usage statisitcs to be measured on the user's end? Possibly with a software solution on their router/proxy server/gateway that measures outgoing requests? What are some keywords or concepts I should search for so that if I receive a request for usage statistics, I can say, "We don't provide usage statistics for practical and logistical reasons but you can roll your own by......."

Aside from how to do it, is providing this service even a good idea? I have heard from a few publishers, that they don't provide this service because it provides justification to those making spending decisions to cancel the subscription. This reasoning is described in better detail here at the American Mathematical Society's website (.pdf). I have also heard from librarians that they hate to cancel subscriptions because they hate, HATE, having gaps in serial publications, not to mention facing the wrath from the 1 or 2 people that do rely on that publication for their work/research.


I would be interested to here about experiences and solutions from the publisher side, as well as the librarian side of things.

posted by chillmost to computers & internet (3 comments total)

If you're on ScienceDirect there's usage statistics here for their journals over the last few years (as a start). If you're in economics try RePEc which automatically generates monthly usage statistics on every article and journal that contributes. Otherwise I'd suggest rolling your own parser on your logs, outputting a .csv file.

As for whether it's a good idea, you won't know until you can compare your stats with other journals that release theirs. If it doesn't look good it may not be a good idea, but I would suggest collecting the statistics for your own benefit anyways.
posted by shmooly at 6:50 AM on September 5, 2008

COUNTER may look complicated, but it's going to be what librarians want, and as a librarian who sometimes has to dig up those stats, I'd strongly suggest going that route. It allows us to compare stats across publications. Otherwise, we get some publishers saying "1000 hits" and they define "hit" as "your user downloaded a pdf of an article" and others say "1000 hits" and they define "hit" as "your user loaded the index of our website".

I'd say releasing stats is better than not. If you're really small, you may also not be expensive enough to be on the cancellation radar unless your stats lead someone to find that literally no one ever accesses your journal from their campus.

Libraries are also going to get other kinds of stats to help evaluate your publications: other "real" stats like click-throughs on a OpenURL link resolver, and other "softer" stats like anecdotal evidence from the reference desk and shouty email from the faculty member who believes your journal is the only decent thing being published today. Actual usage stats will likely help your cause more than they'll hurt it, when combined with that other information, because they give you a chance to say "hey look, here are ACTUAL NUMBERS that justify that money you spent last year".
posted by donnagirl at 7:41 AM on September 5, 2008

2nd donnagirl.

As a librarian I'm usually digging up stats to prove your worth (usage) to my administration for budgetary purposes. It is easier to justify keeping something that I can prove is being used over something that provides no stats. I recently declined to subscribe to an online journal because it provided no stats.
posted by cestmoi15 at 11:28 AM on September 8, 2008